**4. Discussion**

In spite of the considerable understanding of the mechanisms leading to mucosal injury [19], the role of environmental factors in CD pathogenesis remains elusive and, consequently, the treatment is far from optimal. Perturbations of the gu<sup>t</sup> microbiota (dysbiosis) and viral infections have been suggested to trigger the first hit of mucosal inflammation [20,21]. In this regard, a series of elegant experiments has provided some mechanistic insights showing that both gluten- and amylase-trypsin inhibitor-derived peptides generated by pathobiont species, such as *Pseudomonas aeriginosa*, are able to disrupt the epithelial barrier, to activate protease-activated receptors-2 signalling, and to recruit intraepithelial lymphocytes in sensitized mice with a susceptible genetic background [22,23]. Nonetheless, gu<sup>t</sup> mucosa may also harbour gluten-degrading bacteria, such as *Rothia* spp. [24] and *Lactobacillus* spp. [25], with the potential to dampen the harmful e ffects of gluten peptides. When considering studies carried out in adulthood CD, the few performed on duodenal mucosa invariably found a decrease in the relative abundance of *Firmicutes* and an increase of *Proteobacteria*, together with changes of microbiota structure and composition [6,26,27]. However, the small sample size and the lack of a control group do not allow any firm conclusions to be drawn. On the other side, in the vast majority of studies carried out in paediatric CD, the analyses were carried out on stool samples [28]. Since this analysis may miss changes associated with duodenal inflammation or may find others not causally related to the disease process, we sought to investigate the structure and composition of gu<sup>t</sup> microbiota directly at the duodenal level, the main site of tissue injury. We found marked alterations of the ecological indices in CD in terms of reduction of bacterial richness and diversity with respect to dyspeptic patients used as controls. Worthy of note, those su ffering from PCD displayed mucosal indices similar to those with active enteropathy (ACD and RCD). This evidence reinforces the hypothesis that a significant shift of the microbiota composition anticipates the development of mucosal lesions [20]. When considering the data obtained in the groups with active enteropathy, we found the lowest values of α-diversity indices in the three bacterial communities of the RCD group. ACD also showed a reduced microbial richness that, together with a *Proteobacteria*-rich microbiota, has been repeatedly associated with chronic inflammatory conditions [29], including CD [6,26,27]. The TCD group shows intermediate values of both α- and β-diversity comprised between those of ACD and C ones, thus confirming that the GFD does not completely restore a healthy microbiota [6,27].

Moving on to the taxonomic analysis, our results definitely confirm the decrease of *Firmicutes* and increase of *Proteobacteria* in the duodenal mucosa of ACD patients, whereas *Bacteroidetes* displayed a mixed pattern, being decreased in ACD and increased in all the other CD groups, mostly in RCD. Remarkably, Wacklin and coworkers identified *Proteobacteria* as the most abundant phylum in biopsies of TCD patients who su ffered from persistent abdominal symptoms [6,27]. It is known that a balanced gu<sup>t</sup> microbiota is capable of inhibiting uncontrolled expansion of *Proteobacteria*, while a bloom of this phylum has been proposed as reflecting an unstable structure [28,29]. Noteworthily, the profiles of the phyla distribution in salivary samples almost completely mirror those found at the mucosal level, except for *Proteobacteria* that were found predominantly increased in RCD, whilst no substantial correspondence with the other consortia was found in fecal samples.

A further interesting point is the enrichment of the genus *Neisseria* (phylum *Proteobacteria*) and the corresponding family *Neisseriaceae* in duodenal biopsies of ACD patients, in agreemen<sup>t</sup> with previous evidence [26]. Our study enlarges the picture since an increased abundance of *Neisseria* spp. is already evident in PCD, reaches its maximum in ACD and RCD, and then lowers in TCD, although without reaching control levels. This evidence reinforces the hypothesis that it represents a disease-triggering factor instead of a pure consequence of intestinal damage. Moreover, the behaviour of the gluten-degradator species *Rothia* [30] in RCD mucosa should be pointed out since its abundance was similar to controls and significantly higher than in the other CD groups. The reason remains elusive, although it is conceivable that a long-lasting presence of undigested gluten peptides in the gu<sup>t</sup> lumen may select those species with a high capability of degrading them. On the other hand, the depletion of this species in PCD might contribute to loss of gluten tolerance since an incomplete digestion generates oligopeptides with high immunogenicity [2].

Finally, in the salivary ecosystem, it emerges that the SR1 phylum rises from a rate of less than 1% in controls and refractory patients to about 5% in PCD, ACD, and TCD. This is a relatively recently described phylum [31], of which the abundance increases in the oral cavity of subjects with periodontal disease [32]. Another particular trend observed was the critical decrease of both *Lechnospiraceae* and *Fusobacteriaceae* in RCD, whereas they were increased in mucosal samples of noncomplicated CD patients. The taxonomic composition of the fecal community also suggests the existence of a pattern specific for RCD if considering the strong increase of *Pasteurellaceae* and, within this family, of the genus *Haemophilus* spp., in accordance with Cheng et al. [33]. Other taxa generally thought to be protective against inflammation, such as *Faecalibacterium prausnitzii* [34], display a boom in RCD. Therefore, despite the obvious limited number of RCD cases, the presence of particular changes to the microbiota composition in both saliva and fecal samples o ffers important opportunities for screening those cases transitioning to RCD. In addition, the recent finding of full recovery of duodenal architecture and clinical picture in a patient su ffering from RCD following fecal microbiota transplantation for *Clostridium di*ffi*cilie* infection strengthens the relevance of our data [35].

Certainly, our work has some limitations, including the limited sample size. However, it should be pointed out that only those patients who agreed to collect all biological samples were enrolled and that PCD and RCD are rare conditions. Furthermore, despite the oral cavity hardly representing the environment at the duodenal level, we found that saliva displays a microbiota structure and composition more similar to mucosal ones than feces. We also aknowledge that dyspepsia does not represent a real control condition; nonetheless a peculiar dysbiosis, largely dominated by the phylum *Proteobacteria*, genus *Neisseria*, clearly emerged in CD, even before the development of enteropathy. Moreover, the presence of an altered community structure not only in ACD but also in TCD clearly points to the need for additional non-dietary therapies. Since these findings are largely retrievable in the oral cavity, salivary analysis seems a useful tool to capture CD specific signatures. Taken together, our data pave the way for larger studies and support the utility of gu<sup>t</sup> microbiota manipulation for preventive and therapeutic purposes in adulthood CD.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2077-0383/9/4/1109/s1, Figure S1: Plot of relative abundances of the five most abundant families retrieved in each sample biotype; Table S1: Taxonomic assignment across sample biotypes (mucosal, salivary and fecal) in the study cohort; Table S2: α-diversity indices for each sample.

**Author Contributions:** R.C., E.C. and F.B. designed the study. G.F.D.L., S.B. and E.C. were responsible for sample handling and for amplicon production and sequencing. S.M., E.S. and R.B. conceived, designed and implemented the metagenomic pipelines. The operational taxonomic unit abundances were calculated by S.P. and B.B. Phylogenetic analysis was done by S.P., G.F.D.L., S.B. and E.C. Bioinformatic analysis was conceived and implemented by E.S., S.M. and R.B., A.S., E.B. and F.B. were responsible for patient selection and for sample collection and storage. A.V. was responsible for histopathology examination. A.P. and R.C. conducted the human leukocyte antigen genotyping. O.P. was responsible for project and resource administration. R.C. and S.P. wrote the manuscript with contributions from E.C., A.S., E.S. and F.B. L.F. and G.R.C. critically revised the article for important intellectual content. All authors agreeed to be personally accountable for the author's own contributions and for ensuring that questions related to the accuracy or integrity of any part of the work, even ones in which the author was not personally involved, are appropriately investigated, resolved, and documented in the literature. All authors have read and agreed to the published version of the manuscript.

**Funding:** This study was funded by a gran<sup>t</sup> from the Fondazione Celiachia (Italy), project number: N. 016\_FC\_2015. This source of funding did not influence the study design; collection, analysis, and interpretation of data; writing of the manuscript; or decision to submit the article for publication.

**Acknowledgments:** The authors wish to thank Susan West for her thorough revision of the English text and doctor Laura Vanelli for performing the flow cytometric analysis of intraepithelial lymphocytes.

**Conflicts of Interest:** The authors declare no conflict of interest *J. Clin. Med.* **2020**, *9*, 1109
